The increasing use of computed tomography (CT) and cone beam computed tomography (CBCT) in oral and maxillofacial imaging has driven the development of deep learning and radiomics applications to assist clinicians in early diagnosis, accurate prognosis prediction, and efficient treatment planning of maxillofacial diseases. This narrative review aimed to provide an up-to-date overview of the current applications of deep learning and radiomics on CT and CBCT for the diagnosis and management of maxillofacial diseases. Based on current evidence, a wide range of deep learning models on CT/CBCT images have been developed for automatic diagnosis, segmentation, and classification of jaw cysts and tumors, cervical lymph node metastasis, salivary gland diseases, temporomandibular (TMJ) disorders, maxillary sinus pathologies, mandibular fractures, and dentomaxillofacial deformities, while CT-/CBCT-derived radiomics applications mainly focused on occult lymph node metastasis in patients with oral cancer, malignant salivary gland tumors, and TMJ osteoarthritis. Most of these models showed high performance, and some of them even outperformed human experts. The models with performance on par with human experts have the potential to serve as clinically practicable tools to achieve the earliest possible diagnosis and treatment, leading to a more precise and personalized approach for the management of maxillofacial diseases. Challenges and issues, including the lack of the generalizability and explainability of deep learning models and the uncertainty in the reproducibility and stability of radiomic features, should be overcome to gain the trust of patients, providers, and healthcare organizers for daily clinical use of these models.

This preliminary study of head and neck tumors demonstrated that amide proton transfer imaging can help differentiate malignant tumors from normal tissues and from a group of benign tumors, but cannot be used to differentiate between different malignant tumor groups. Purpose To evaluate the utility of amide proton transfer (APT) imaging in the characterization of head and neck tumors. Materials and Methods This retrospective study of APT imaging included 117 patients with 70 nasopharyngeal undifferentiated carcinomas (NUCs), 26 squamous cell carcinomas (SCCs), eight non-Hodgkin lymphomas (NHLs), and 13 benign salivary gland tumors (BSGTs). Normal tissues were examined in 25 patients. The APT means of malignant tumors, normal tissues, and benign tumors were calculated and compared with the Student t test and analysis of variance. The added value of the mean APT to the mean apparent diffusion coefficient (ADC) for differentiating malignant and benign tumors was evaluated by using receiver operating characteristic analysis and integrated discrimination index. Results The mean APT of malignant tumors (2.40% ± 0.97 [standard deviation]) was significantly higher than that of brain tissue (1.13% ± 0.43), muscle tissue (0.23% ± 0.73), and benign tumors (1.32% ± 1.20) (P < .00...

BACKGROUND AND PURPOSE: MR imaging can detect nasopharyngeal carcinoma that is hidden from endoscopic view, but for accurate detection carcinoma confined within the nasopharynx (stage T1) must be distinguished from benign hyperplasia of the nasopharynx. This study aimed to document the MR imaging features of stage T1 nasopharyngeal carcinoma and to attempt to identify features distinguishing it from benign hyperplasia. MATERIALS AND METHODS: MR images of 189 patients with nasopharyngeal carcinoma confined to the nasopharynx and those of 144 patients with benign hyperplasia were reviewed and compared in this retrospective study. The center, volume, size asymmetry (maximum percentage difference in area between the right and left nasopharyngeal halves), signal intensity asymmetry, deep mucosal white line (greater contrast enhancement along the deep tumor margin), and absence/distortion of the adenoidal septa were evaluated. Differences were assessed with logistic regression and the χ² test. RESULTS: The nasopharyngeal carcinoma center was lateral, central, or diffuse in 134/189 (70.9%), 25/189 (13.2%), and 30/189 (15.9%) cases, respectively. Nasopharyngeal carcinomas involving the walls showed that a deep mucosal white line was present in 180/183 (98.4%), with a focal loss of this line in 153/180 (85%) cases. Adenoidal septa were absent or distorted in 111/111 (100%) nasopharyngeal carcinomas involving the adenoid. Compared with benign hyperplasia, nasopharyngeal carcinoma had a significantly greater volume, size asymmetry, signal asymmetry, focal loss of the deep mucosal white line, and absence/distortion of the adenoidal septa (P < .001). Although size asymmetry was the most accurate criterion (89.5%) for nasopharyngeal carcinoma detection, use of this parameter alone would have missed 11.9% of early-stage T1 nasopharyngeal carcinomas. CONCLUSIONS: MR imaging features can help distinguish stage T1 nasopharyngeal carcinoma from benign hyperplasia in most cases.

BACKGROUND AND PURPOSE: Pretreatment prediction of patients with nasopharyngeal carcinoma who will fail conventional treatment would potentially allow these patients to undergo more intensive treatment or closer posttreatment monitoring. The aim of the study was to determine the ability of pretreatment DWI to predict local failure in patients with nasopharyngeal carcinoma based on long-term clinical outcome. MATERIALS AND METHODS: One hundred fifty-eight patients with pretreatment DWI underwent analysis of the primary tumor to obtain the ADC mean, ADC skewness, ADC kurtosis, volume, and T-stage. Univariate and multivariate analyses using logistic regression were performed to compare the ADC parameters, volume, T-stage, and patient age in primary tumors with local failure and those with local control, by using a minimum of 5-year follow-up to confirm local control. RESULTS: Local control was achieved in 131/158 (83%) patients (range, 60.3–117.7 months) and local failure occurred in 27/158 (17%) patients (range, 5.2–79.8 months). Compared with tumors with local control, those with local failure showed a significantly lower ADC skewness (ADC values with the greatest frequencies were shifted away from the lower ADC range) (P = .006) and lower ADC kurtosis (curve peak broader) (P = .024). The ADC skewness remained significant on multivariate analysis (P = .044). There was a trend toward higher tumor volumes in local failure, but the volume, together with T-stage and ADC mean, were not significantly different between the 2 groups. CONCLUSIONS: Pretreatment DWI of primary tumors found that the skewness of the ADC distribution curve was a predictor of local failure in patients with nasopharyngeal carcinoma, based on long-term clinical outcome.

Objectives: This study aimed to systematically review the current performance of large language models (LLMs) in dento-maxillofacial radiology (DMFR). Methods: Five electronic databases were used to identify studies that developed, fine-tuned, or evaluated LLMs for DMFR-related tasks. Data extracted included study purpose, LLM type, images/text source, applied language, dataset characteristics, input and output, performance outcomes, evaluation methods, and reference standards. Customized assessment criteria adapted from the TRIPOD-LLM reporting guideline were used to evaluate the risk-of-bias in the included studies specifically regarding the clarity of dataset origin, the robustness of performance evaluation methods, and the validity of the reference standards. Results: The initial search yielded 1621 titles, and 19 studies were included. These studies investigated the use of LLMs for tasks including the production and answering of DMFR-related qualification exams and educational questions (n = 8), diagnosis and treatment recommendations (n = 7), and radiology report generation and patient communication (n = 4). LLMs demonstrated varied performance in diagnosing dental conditions, with accuracy ranging from 37% to 92.5% and expert ratings for differential diagnosis and treatment planning between 3.6 and 4.7 on a 5-point scale. For DMFR-related qualification exams and board-style questions, LLMs achieved correctness rates between 33.3% and 86.1%. Automated radiology report generation showed moderate performance with accuracy ranging from 70.4% to 81.3%. Conclusions: LLMs demonstrate promising potential in DMFR, particularly for diagnostic, educational, and report generation tasks. However, their current accuracy, completeness, and consistency remain variable. Further development, validation, and standardization are needed before LLMs can be reliably integrated as supportive tools in clinical workflows and educational settings.